Inductive Inference and Beyond: From Bayesian
Rationality to
Logical Reliability
Prasanta S. Bandyopadhyay
Montana State University
Traditional epistemology fails to provide a satisfactory account of inductive inference
because the notion of justification raises a host of problems. One well-known approach to
induction is Bayesianism that exploits the concept of rationality. Bayesianism is,
however, alleged to rest on dubious assumptions. As an alternative, logical reliabilism
(or Android epistemology) defended by Glymour, Kelly, and Juhl provides an account of
induction that draws its inspiration from computability theory and the formal theory of
learning. Logical reliabilists contend that there are similarities between induction and
computability theory. In addition, formal learning theory provides necessary tools for
them to develop algorithms for key episodes in the history of science.
I develop and defend Bayesianism. I contend that logical reliabilism takes for granted certain questionable assumptions, too. The important question is which assumptions are more reasonable. I also argue that my approach provides a more plausible and more intuitive reconstruction of some episodes of science than theirs.